Merge branch 'feature/py3_for_gen_esp_err_to_name' into 'master'

tools: support python3 in gen_esp_err_to_name.py

See merge request idf/esp-idf!2827
This commit is contained in:
Angus Gratton 2018-08-27 12:29:03 +08:00
commit 2c8c210a66
3 changed files with 46 additions and 26 deletions

View File

@ -362,10 +362,21 @@ test_idf_monitor:
test_esp_err_to_name_on_host:
<<: *host_test_template
artifacts:
when: on_failure
paths:
- components/esp32/esp_err_to_name.c
expire_in: 1 week
script:
- cd tools/
- source /opt/pyenv/activate
- pyenv global 2.7.15
- ./gen_esp_err_to_name.py
- git diff --exit-code -- ../components/esp32/esp_err_to_name.c || (echo 'Differences found. Please run gen_esp_err_to_name.py and commit the changes.'; exit 1)
- pyenv global 3.4.8
- ./gen_esp_err_to_name.py
- git diff --exit-code -- ../components/esp32/esp_err_to_name.c || (echo 'Differences found between running under Python 2 and 3.'; exit 1)
- pyenv global system
push_to_github:
stage: deploy

View File

@ -125,11 +125,11 @@ static const esp_err_msg_t esp_err_msg_table[] = {
ERR_TBL_IT(ESP_ERR_NVS_INVALID_HANDLE), /* 4359 0x1107 Handle has been closed or is NULL */
# endif
# ifdef ESP_ERR_NVS_REMOVE_FAILED
ERR_TBL_IT(ESP_ERR_NVS_REMOVE_FAILED), /* 4360 0x1108 The value wasnt updated because flash
write operation has failed. The value was
written however, and update will be finished
after re-initialization of nvs, provided
that flash operation doesnt fail again. */
ERR_TBL_IT(ESP_ERR_NVS_REMOVE_FAILED), /* 4360 0x1108 The value wasnt updated because flash write
operation has failed. The value was written
however, and update will be finished after
re-initialization of nvs, provided that
flash operation doesnt fail again. */
# endif
# ifdef ESP_ERR_NVS_KEY_TOO_LONG
ERR_TBL_IT(ESP_ERR_NVS_KEY_TOO_LONG), /* 4361 0x1109 Key name is too long */

View File

@ -14,13 +14,19 @@
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import print_function
from __future__ import unicode_literals
from builtins import str
from builtins import range
from builtins import object
from io import open
import os
import argparse
import re
import fnmatch
import string
import collections
import textwrap
import functools
# list files here which should not be parsed
ignore_files = [ 'components/mdns/test_afl_fuzz_host/esp32_compat.h' ]
@ -35,7 +41,7 @@ err_dict = collections.defaultdict(list) #identified errors are stored here; map
rev_err_dict = dict() #map of error string to error code
unproc_list = list() #errors with unknown codes which depend on other errors
class ErrItem:
class ErrItem(object):
"""
Contains information about the error:
- name - error string
@ -99,7 +105,7 @@ def process(line, idf_path):
words = re.split(r' +', line, 2)
# words[1] is the error name
# words[2] is the rest of the line (value, base + value, comment)
if len(words) < 2:
if len(words) < 3:
raise InputError(idf_path, "Error at line %s" % line)
line = ""
@ -109,8 +115,8 @@ def process(line, idf_path):
# identify possible comment
m = re.search(r'/\*!<(.+?(?=\*/))', todo_str)
if m:
comment = string.strip(m.group(1))
todo_str = string.strip(todo_str[:m.start()]) # keep just the part before the comment
comment = m.group(1).strip()
todo_str = todo_str[:m.start()].strip() # keep just the part before the comment
# identify possible parentheses ()
m = re.search(r'\((.+)\)', todo_str)
@ -181,14 +187,14 @@ def path_to_include(path):
are inside the "include" directory. Other special cases need to be handled
here when the compiler gives an unknown header file error message.
"""
spl_path = string.split(path, os.sep)
spl_path = path.split(os.sep)
try:
i = spl_path.index('include')
except ValueError:
# no include in the path -> use just the filename
return os.path.basename(path)
else:
return str(os.sep).join(spl_path[i+1:]) # subdirectories and filename in "include"
return os.sep.join(spl_path[i+1:]) # subdirectories and filename in "include"
def print_warning(error_list, error_code):
"""
@ -200,7 +206,7 @@ def print_warning(error_list, error_code):
def max_string_width():
max = 0
for k in err_dict.keys():
for k in err_dict:
for e in err_dict[k]:
x = len(e.name)
if x > max:
@ -214,7 +220,7 @@ def generate_c_output(fin, fout):
"""
# make includes unique by using a set
includes = set()
for k in err_dict.keys():
for k in err_dict:
for e in err_dict[k]:
includes.add(path_to_include(e.file))
@ -226,7 +232,7 @@ def generate_c_output(fin, fout):
include_list.sort()
max_width = max_string_width() + 17 + 1 # length of " ERR_TBL_IT()," with spaces is 17
max_decdig = max(len(str(k)) for k in err_dict.keys())
max_decdig = max(len(str(k)) for k in err_dict)
for line in fin:
if re.match(r'@COMMENT@', line):
@ -239,7 +245,7 @@ def generate_c_output(fin, fout):
last_file = ""
for k in sorted(err_dict.keys()):
if len(err_dict[k]) > 1:
err_dict[k].sort()
err_dict[k].sort(key=functools.cmp_to_key(ErrItem.__cmp__))
print_warning(err_dict[k], k)
for e in err_dict[k]:
if e.file != last_file:
@ -297,22 +303,25 @@ def main():
path_in_idf = os.path.relpath(full_path, idf_path)
if path_in_idf in ignore_files or path_in_idf.startswith(ignore_dirs):
continue
with open(full_path, "r+") as f:
for line in f:
# match also ESP_OK and ESP_FAIL because some of ESP_ERRs are referencing them
if re.match(r"\s*#define\s+(ESP_ERR_|ESP_OK|ESP_FAIL)", line):
try:
process(str.strip(line), path_in_idf)
except InputError as e:
print (e)
with open(full_path, encoding='utf-8') as f:
try:
for line in f:
# match also ESP_OK and ESP_FAIL because some of ESP_ERRs are referencing them
if re.match(r"\s*#define\s+(ESP_ERR_|ESP_OK|ESP_FAIL)", line):
try:
process(line.strip(), path_in_idf)
except InputError as e:
print (e)
except UnicodeDecodeError:
raise ValueError("The encoding of {} is not Unicode.".format(path_in_idf))
process_remaining_errors()
if args.rst_output is not None:
with open(args.rst_output, 'w') as fout:
with open(args.rst_output, 'w', encoding='utf-8') as fout:
generate_rst_output(fout)
else:
with open(args.c_input, 'r') as fin, open(args.c_output, 'w') as fout:
with open(args.c_input, 'r', encoding='utf-8') as fin, open(args.c_output, 'w', encoding='utf-8') as fout:
generate_c_output(fin, fout)
if __name__ == "__main__":